- My Forums
- Tiger Rant
- LSU Recruiting
- SEC Rant
- Saints Talk
- Pelicans Talk
- More Sports Board
- Fantasy Sports
- Golf Board
- Soccer Board
- O-T Lounge
- Tech Board
- Home/Garden Board
- Outdoor Board
- Health/Fitness Board
- Movie/TV Board
- Book Board
- Music Board
- Political Talk
- Money Talk
- Fark Board
- Gaming Board
- Travel Board
- Food/Drink Board
- Ticket Exchange
- TD Help Board
Customize My Forums- View All Forums
- Show Left Links
- Topic Sort Options
- Trending Topics
- Recent Topics
- Active Topics
Started By
Message
Team Talent by Class (for Analytical Nerds)
Posted on 12/28/20 at 6:07 pm
Posted on 12/28/20 at 6:07 pm
Here’s the TL;DR version of this post:
The recent talk about LSU's attrition from the 2019 team, along with the impact of the less-than-stellar 2018 recruiting class, got me wondering: what does LSU's talent by class look like compared to other SEC teams? Is LSU in a "hole" for upperclassmen compared to the rest of the league? Let's find out.
First I recreated the 247Sports Team Talent Composite using roster data from each SEC team, corrected to account for some transfers/opt-outs/dismissals/etc. that weren’t captured in the 247Sports data:
Next, I looked at how each class contributed to the total for each team:
And here are the raw totals for each individual class, ranked highest to lowest. Note that these utilize a different value for sigma than the overall team talent composite since they are smaller sample sizes. In layman’s terms, the score for each "class" below is effectively the same as the 247 class score would be for a recruiting class including that group of players:
A couple of immediate observations:
1. Every team except Vanderbilt sees their score drop from the freshman class to the senior class. This makes sense, logically. All schools battle attrition - guys get processed, drafted, graduate, transfer, etc. - and if you take 25 new freshmen every year, your freshman class will always be the largest.
2. Georgia and Alabama lead the conference in terms of talent for every class. In fact, their junior and senior classes account for about half of the talent gap with the rest of the conference. In other words - they are retaining talented upperclassmen. Ouch.
3. LSU's freshman class looks pretty damned good. The junior class was about where I expected it to be - you can definitely see the hole that 2018 left. The surprising thing to me was the drop off from the freshman to sophomore class.
Now, it's worth noting that the class of 2018 contributes to sophomores and juniors for all of these teams. That's because a freshman who started in 2018 can potentially be either a RS sophomore or a "true" junior this year, depending on playing time. However, there's another factor. Five of our top thirteen recruits from the 2019 class are no longer with the team due to either transfers or dismissals. That is, quite honestly, abysmal roster turnover.
4. LSU’s senior class actually looks better than I expected, for some reason.
Lastly, let’s compare LSU’s talent for each class to the SEC median:
The SEC median is represented by 0% / the dotted purple line. Bars to the right mean that LSU’s talent exceeds the SEC median. Bars to the left mean that our talent is less than the SEC median. The first bar on this graph is the total roster talent – again you can see that LSU is +7.89% above the SEC median in this category. The remaining bars are the individual class ratings. You can see that our junior class is rated below the SEC median, while the rest of the roster exceeds the SEC median.
My biggest takeaway here is that LSU may or may not have a massive attrition problem relative to the rest of the league. In other words – the question (to me) that we need to answer is whether the 2018 class attrition was a one-off, or part of a much larger problem. In my opinion, the data is a bit inconclusive in that respect. The 2018 class is a big part of our below-average junior class, and it’s part of the reason for our slightly below-average (compared to the school’s total) sophomore class. The senior class, on the other hand, was likely heavily impacted by the fact that we won a national title last year.
DISCLAIMER: If you aren't interested in statistics or analytics, you will probably find the following post(s) to be long and boring. If that's the case, stop reading now or continue at your own risk. I take no responsibility for the massive waste of time that will surely come out of your reading the remainder of this diatribe.
The recent talk about LSU's attrition from the 2019 team, along with the impact of the less-than-stellar 2018 recruiting class, got me wondering: what does LSU's talent by class look like compared to other SEC teams? Is LSU in a "hole" for upperclassmen compared to the rest of the league? Let's find out.
First I recreated the 247Sports Team Talent Composite using roster data from each SEC team, corrected to account for some transfers/opt-outs/dismissals/etc. that weren’t captured in the 247Sports data:
Next, I looked at how each class contributed to the total for each team:
And here are the raw totals for each individual class, ranked highest to lowest. Note that these utilize a different value for sigma than the overall team talent composite since they are smaller sample sizes. In layman’s terms, the score for each "class" below is effectively the same as the 247 class score would be for a recruiting class including that group of players:
A couple of immediate observations:
1. Every team except Vanderbilt sees their score drop from the freshman class to the senior class. This makes sense, logically. All schools battle attrition - guys get processed, drafted, graduate, transfer, etc. - and if you take 25 new freshmen every year, your freshman class will always be the largest.
2. Georgia and Alabama lead the conference in terms of talent for every class. In fact, their junior and senior classes account for about half of the talent gap with the rest of the conference. In other words - they are retaining talented upperclassmen. Ouch.
3. LSU's freshman class looks pretty damned good. The junior class was about where I expected it to be - you can definitely see the hole that 2018 left. The surprising thing to me was the drop off from the freshman to sophomore class.
Now, it's worth noting that the class of 2018 contributes to sophomores and juniors for all of these teams. That's because a freshman who started in 2018 can potentially be either a RS sophomore or a "true" junior this year, depending on playing time. However, there's another factor. Five of our top thirteen recruits from the 2019 class are no longer with the team due to either transfers or dismissals. That is, quite honestly, abysmal roster turnover.
4. LSU’s senior class actually looks better than I expected, for some reason.
Lastly, let’s compare LSU’s talent for each class to the SEC median:
The SEC median is represented by 0% / the dotted purple line. Bars to the right mean that LSU’s talent exceeds the SEC median. Bars to the left mean that our talent is less than the SEC median. The first bar on this graph is the total roster talent – again you can see that LSU is +7.89% above the SEC median in this category. The remaining bars are the individual class ratings. You can see that our junior class is rated below the SEC median, while the rest of the roster exceeds the SEC median.
My biggest takeaway here is that LSU may or may not have a massive attrition problem relative to the rest of the league. In other words – the question (to me) that we need to answer is whether the 2018 class attrition was a one-off, or part of a much larger problem. In my opinion, the data is a bit inconclusive in that respect. The 2018 class is a big part of our below-average junior class, and it’s part of the reason for our slightly below-average (compared to the school’s total) sophomore class. The senior class, on the other hand, was likely heavily impacted by the fact that we won a national title last year.
DISCLAIMER: If you aren't interested in statistics or analytics, you will probably find the following post(s) to be long and boring. If that's the case, stop reading now or continue at your own risk. I take no responsibility for the massive waste of time that will surely come out of your reading the remainder of this diatribe.
This post was edited on 12/28/20 at 6:13 pm
21
Posted on 12/28/20 at 6:08 pm to lostinbr
DISCLAIMER: I gave you a TL;DR, dammit. Don’t say I didn’t warn you!
The recent talk about LSU's attrition from the 2019 team, along with the impact of the less-than-stellar 2018 recruiting class, got me wondering: what does LSU's talent by class look like compared to other SEC teams? Is LSU in a "hole" for upperclassmen compared to the rest of the league? Let's find out.
First, we need to define roster talent. This is the easy part - I used the 247 Team Talent Composite Rankings as a guideline. For those of you who are unfamiliar, the 247 Team Talent Composite is a system for ranking the amount of, well, talent on each team's roster for a given season based on each player's 247Sports Composite Rating as a high school recruit. For reference, here's what the Team Talent Rankings look like for the 2020 season:
Unfortunately, 247Sports doesn’t offer a downloadable database with their recruiting data. So the only way to re-engineer the Team Talent Composite from their website is to manually enter the ratings for each player on each team. frick that noise.
But we're in luck! CollegeFootballData.com has tons of downloadable data, which they scrape from various web sources, for the football nerd to play with. Unfortunately (again) they don't have a single database of rosters with high school recruiting ratings. They do, however, have a database of rosters and a separate database of 247 Composite recruiting ratings. For various reasons that you can probably imagine, matching up the recruiting database with the roster database was really the hardest part of this entire analysis. I had to manually enter the ratings for many players because my database lookup ran into some complications. More on that later.
For now, though, let's assume we have a valid database of team rosters with 247Sports Composite recruiting ratings for each player. Now what? Well, first I wanted to make sure I could actually recreate the 247Sports Team Talent Composite ratings. Otherwise, I'm just barking up the wrong tree. Here's the formula from 247Sports' website:
There's a much longer explanation of how the rating system works on their website, but here's the gist of it: they add up the individual player ratings on a roster, but each player rating is weighted. The best player on the roster counts for 100% of their rating. The second-best player counts for slightly less, and so on until you reach a point where additional players are basically a rounding error. The idea is that players can only contribute if they see the field, and your best players are the most likely to play. There's a huge difference between having 20 4-star guys and having 30 4-star guys. However, there's less of a difference between 30 and 40, and even less of a difference between 40 and 50. It's a system that makes sense logically, and they use the same system with a different parameter for sigma in their recruiting class rankings (more on that later as well).
A couple of additional comments about the 247Sports Team Talent Composite formula:
- The description says "Rn" is equal to the recruit's rating x 100. Meaning a player with a 0.9760 rating would have Rn = 97.60. However, it turns out this isn't actually how they do things. They actually use a 30-point scale with the formula Rn = (rating x 100) - 70. So that 0.9760 recruit actually has Rn = 27.60. (Thanks, Reddit.)
- I have no earthly idea how they figure out their value for sigma. Supposedly it's a standard deviation of some sort but I sure as hell couldn't replicate it. Instead, I used a couple of sample teams and back-calculated sigma from the actual team talent ratings.
This is about the time that I realized the severity of my data problem. As I went through my database manually adding/editing various players' ratings, I realized something: my database wasn't the only tool that had issues. The rosters listed on the 247Sports Team Talent Composite also have problems. For starters, they often exclude preferred walk-ons even if they actually had a 247Sports Composite rating out of high school. Same for transfers in some cases. Also, they haven't done a great job of keeping up with opt-outs and players entering the transfer portal this year. So I came up with a set of rules:
1. If a player was listed both on my roster data (from CollegeFootballData, which scrapes from various web sources) AND the 247Sports Team Talent Composite roster data, I included their rating in my calcs.
2. If a player was missing from one of the two sources, I verified manually whether they were still on the team. If they had opted out, entered the transfer portal, or left the team for other reasons, I excluded them from the calcs. If they were in fact still on the team, I included them.
Ultimately I wound up having to go through and verify this for every SEC team. I would have liked to have cast a wider net (outside the SEC for example) for this analysis, but frick all that. It's way too much work.
Here's what I came up with in the end for team talent:
I verified that if I match the 247Sports rosters exactly, I get the exact same results as them. My actual results are slightly different due to the opt-outs & transfers described previously. But I actually think they're more accurate since I verified them manually.
Note that all of the data from this point on includes Arik Gilbert and Terrace Marshall for the LSU calcs. I was torn on this, but elected to leave them in the calcs since both played a majority of the season.
ETA: I addressed the Marshall/Gilbert losses in this post on page 3.
The recent talk about LSU's attrition from the 2019 team, along with the impact of the less-than-stellar 2018 recruiting class, got me wondering: what does LSU's talent by class look like compared to other SEC teams? Is LSU in a "hole" for upperclassmen compared to the rest of the league? Let's find out.
First, we need to define roster talent. This is the easy part - I used the 247 Team Talent Composite Rankings as a guideline. For those of you who are unfamiliar, the 247 Team Talent Composite is a system for ranking the amount of, well, talent on each team's roster for a given season based on each player's 247Sports Composite Rating as a high school recruit. For reference, here's what the Team Talent Rankings look like for the 2020 season:
Unfortunately, 247Sports doesn’t offer a downloadable database with their recruiting data. So the only way to re-engineer the Team Talent Composite from their website is to manually enter the ratings for each player on each team. frick that noise.
But we're in luck! CollegeFootballData.com has tons of downloadable data, which they scrape from various web sources, for the football nerd to play with. Unfortunately (again) they don't have a single database of rosters with high school recruiting ratings. They do, however, have a database of rosters and a separate database of 247 Composite recruiting ratings. For various reasons that you can probably imagine, matching up the recruiting database with the roster database was really the hardest part of this entire analysis. I had to manually enter the ratings for many players because my database lookup ran into some complications. More on that later.
For now, though, let's assume we have a valid database of team rosters with 247Sports Composite recruiting ratings for each player. Now what? Well, first I wanted to make sure I could actually recreate the 247Sports Team Talent Composite ratings. Otherwise, I'm just barking up the wrong tree. Here's the formula from 247Sports' website:
There's a much longer explanation of how the rating system works on their website, but here's the gist of it: they add up the individual player ratings on a roster, but each player rating is weighted. The best player on the roster counts for 100% of their rating. The second-best player counts for slightly less, and so on until you reach a point where additional players are basically a rounding error. The idea is that players can only contribute if they see the field, and your best players are the most likely to play. There's a huge difference between having 20 4-star guys and having 30 4-star guys. However, there's less of a difference between 30 and 40, and even less of a difference between 40 and 50. It's a system that makes sense logically, and they use the same system with a different parameter for sigma in their recruiting class rankings (more on that later as well).
A couple of additional comments about the 247Sports Team Talent Composite formula:
- The description says "Rn" is equal to the recruit's rating x 100. Meaning a player with a 0.9760 rating would have Rn = 97.60. However, it turns out this isn't actually how they do things. They actually use a 30-point scale with the formula Rn = (rating x 100) - 70. So that 0.9760 recruit actually has Rn = 27.60. (Thanks, Reddit.)
- I have no earthly idea how they figure out their value for sigma. Supposedly it's a standard deviation of some sort but I sure as hell couldn't replicate it. Instead, I used a couple of sample teams and back-calculated sigma from the actual team talent ratings.
This is about the time that I realized the severity of my data problem. As I went through my database manually adding/editing various players' ratings, I realized something: my database wasn't the only tool that had issues. The rosters listed on the 247Sports Team Talent Composite also have problems. For starters, they often exclude preferred walk-ons even if they actually had a 247Sports Composite rating out of high school. Same for transfers in some cases. Also, they haven't done a great job of keeping up with opt-outs and players entering the transfer portal this year. So I came up with a set of rules:
1. If a player was listed both on my roster data (from CollegeFootballData, which scrapes from various web sources) AND the 247Sports Team Talent Composite roster data, I included their rating in my calcs.
2. If a player was missing from one of the two sources, I verified manually whether they were still on the team. If they had opted out, entered the transfer portal, or left the team for other reasons, I excluded them from the calcs. If they were in fact still on the team, I included them.
Ultimately I wound up having to go through and verify this for every SEC team. I would have liked to have cast a wider net (outside the SEC for example) for this analysis, but frick all that. It's way too much work.
Here's what I came up with in the end for team talent:
I verified that if I match the 247Sports rosters exactly, I get the exact same results as them. My actual results are slightly different due to the opt-outs & transfers described previously. But I actually think they're more accurate since I verified them manually.
Note that all of the data from this point on includes Arik Gilbert and Terrace Marshall for the LSU calcs. I was torn on this, but elected to leave them in the calcs since both played a majority of the season.
ETA: I addressed the Marshall/Gilbert losses in this post on page 3.
This post was edited on 12/29/20 at 12:17 pm
Posted on 12/28/20 at 6:08 pm to lostinbr
So now I know my data and calcs are working - what does the breakdown look like by class for each team? I ran into another roadblock here. As it turns out, that value for sigma that I mentioned earlier matters. In essence it determines where the inflection point is for players to start being weighted more heavily. If I used the same sigma value for each individual class (freshmen, sophomores, etc.) that I used for the entire roster, it caused players to be weighted way to heavily IMO.
After taking a step back, I came to a solution: I used the sigma value from the 247Sports single-year recruiting rankings. In other words, the score for each "class" below is effectively the same as the 247 class score would be for a recruiting class including that group of players.
This first graph is actually normalized to the overall team talent scores presented earlier (except the Y-axis actually goes all the way to zero). The idea is that this should give you an idea of the each class's contribution to each team's overall score:
And here are the raw totals for each individual class, ranked highest to lowest:
A couple of immediate observations:
1. Every team except Vanderbilt sees their score drop from the freshman class to the senior class. This makes sense, logically. All schools battle attrition - guys get processed, drafted, graduate, transfer, etc. - and if you take 25 new freshmen every year, your freshman class will always be the largest.
2. Georgia and Alabama lead the conference in terms of talent for every class. In fact, their junior and senior classes account for about half of the talent gap with the rest of the conference. In other words - they are retaining talented upperclassmen. Ouch.
3. LSU's freshman class looks pretty damned good. The junior class was about where I expected it to be - you can definitely see the hole that 2018 left. The surprising thing to me was the drop off from the freshman to sophomore class.
Now, it's worth noting that the class of 2018 contributes to sophomores and juniors for all of these teams. That's because a freshman who started in 2018 can potentially be either a RS sophomore or a "true" junior this year, depending on playing time. However, there's another factor. Five of our top thirteen recruits from the 2019 class are no longer with the team due to either transfers or dismissals. That is, quite honestly, abysmal roster turnover.
4. LSU’s senior class actually looks better than I expected, for some reason.
Lastly, let’s compare LSU’s talent for each class to the SEC median:
The SEC median is represented by 0% / the dotted purple line. Bars to the right mean that LSU’s talent exceeds the SEC median. Bars to the left mean that our talent is less than the SEC median. The first bar on this graph is the total roster talent – again you can see that LSU is +7.89% above the SEC median in this category. The remaining bars are the individual class ratings. You can see that our junior class is rated below the SEC median, while the rest of the roster exceeds the SEC median.
My biggest takeaway here is that LSU may or may not have a massive attrition problem relative to the rest of the league. In other words – the question (to me) that we need to answer is whether the 2018 class attrition was a one-off, or part of a much larger problem. In my opinion, the data is a bit inconclusive in that respect. The 2018 class is a big part of our below-average junior class, and it’s part of the reason for our slightly below-average (compared to the school’s total) sophomore class. The senior class, on the other hand, was likely heavily impacted by the fact that we won a national title last year.
Which means I really haven’t accomplished much other than wasting a shitload of time putting all of this together. But maybe I’ll take a deeper dive into each team’s attrition on another day!
In summary, TulaneLSU ain't got shite on me!
After taking a step back, I came to a solution: I used the sigma value from the 247Sports single-year recruiting rankings. In other words, the score for each "class" below is effectively the same as the 247 class score would be for a recruiting class including that group of players.
This first graph is actually normalized to the overall team talent scores presented earlier (except the Y-axis actually goes all the way to zero). The idea is that this should give you an idea of the each class's contribution to each team's overall score:
And here are the raw totals for each individual class, ranked highest to lowest:
A couple of immediate observations:
1. Every team except Vanderbilt sees their score drop from the freshman class to the senior class. This makes sense, logically. All schools battle attrition - guys get processed, drafted, graduate, transfer, etc. - and if you take 25 new freshmen every year, your freshman class will always be the largest.
2. Georgia and Alabama lead the conference in terms of talent for every class. In fact, their junior and senior classes account for about half of the talent gap with the rest of the conference. In other words - they are retaining talented upperclassmen. Ouch.
3. LSU's freshman class looks pretty damned good. The junior class was about where I expected it to be - you can definitely see the hole that 2018 left. The surprising thing to me was the drop off from the freshman to sophomore class.
Now, it's worth noting that the class of 2018 contributes to sophomores and juniors for all of these teams. That's because a freshman who started in 2018 can potentially be either a RS sophomore or a "true" junior this year, depending on playing time. However, there's another factor. Five of our top thirteen recruits from the 2019 class are no longer with the team due to either transfers or dismissals. That is, quite honestly, abysmal roster turnover.
4. LSU’s senior class actually looks better than I expected, for some reason.
Lastly, let’s compare LSU’s talent for each class to the SEC median:
The SEC median is represented by 0% / the dotted purple line. Bars to the right mean that LSU’s talent exceeds the SEC median. Bars to the left mean that our talent is less than the SEC median. The first bar on this graph is the total roster talent – again you can see that LSU is +7.89% above the SEC median in this category. The remaining bars are the individual class ratings. You can see that our junior class is rated below the SEC median, while the rest of the roster exceeds the SEC median.
My biggest takeaway here is that LSU may or may not have a massive attrition problem relative to the rest of the league. In other words – the question (to me) that we need to answer is whether the 2018 class attrition was a one-off, or part of a much larger problem. In my opinion, the data is a bit inconclusive in that respect. The 2018 class is a big part of our below-average junior class, and it’s part of the reason for our slightly below-average (compared to the school’s total) sophomore class. The senior class, on the other hand, was likely heavily impacted by the fact that we won a national title last year.
Which means I really haven’t accomplished much other than wasting a shitload of time putting all of this together. But maybe I’ll take a deeper dive into each team’s attrition on another day!
In summary, TulaneLSU ain't got shite on me!
This post was edited on 12/28/20 at 6:13 pm
Posted on 12/28/20 at 6:10 pm to lostinbr
holy graphs.. good stuff tho
Posted on 12/28/20 at 6:11 pm to lostinbr
Posted on 12/28/20 at 6:12 pm to lostinbr
I just want to know if we're going to be good next year?
Posted on 12/28/20 at 6:12 pm to lostinbr
To answer...I would say it's about 350.
Posted on 12/28/20 at 6:12 pm to DBG
Posted on 12/28/20 at 6:16 pm to lostinbr
Just mind boggling that LSU lost to Missouri.
Posted on 12/28/20 at 6:16 pm to lostinbr
This is going to take some time to digest but kudos for putting this all together.
Posted on 12/28/20 at 6:18 pm to lostinbr
Thank you lostinbr. I’m a visual person myself.
Imo more than anything this shows how even one down class can affect the entire roster management task. A lot of 2018 commits transferred out because they knew they’d never see the field. Also 2018 included a decent amount of JUCO and grad transfers who only have a year or two of eligibility.
On the other hand, hitting on 3 straight classes has set LSU up to win or at least compete for a title in 2022.
Imo more than anything this shows how even one down class can affect the entire roster management task. A lot of 2018 commits transferred out because they knew they’d never see the field. Also 2018 included a decent amount of JUCO and grad transfers who only have a year or two of eligibility.
On the other hand, hitting on 3 straight classes has set LSU up to win or at least compete for a title in 2022.
Posted on 12/28/20 at 6:19 pm to jbraua
quote:
I just want to know if we're going to be good next year?
My best guess is that, from a pure roster talent standpoint, we will be neck-and-neck with aTm and UF for that #3 spot. That obviously depends on who stays and who goes at each school.
One key piece that is missing in that evaluation though - aTm and UF could both lose their quarterbacks this offseason, and whoever winds up QB1 at LSU will have starting experience (unless Nuss blows everyone away).
Posted on 12/28/20 at 6:20 pm to lostinbr
Good stuff.
Unfortunately my biggest takeaway is that we can't rely on talent alone to continue beating A&M. Instead we're going to have to hire assistants smart enough to go head to head with Fisher.
Unfortunately my biggest takeaway is that we can't rely on talent alone to continue beating A&M. Instead we're going to have to hire assistants smart enough to go head to head with Fisher.
Posted on 12/28/20 at 6:21 pm to lostinbr
Posted on 12/28/20 at 6:21 pm to goodgrin
quote:
Just mind boggling that LSU lost to Missouri.
And Miss St
I appreciate the OP for the work he put in. Nice work.
This post was edited on 12/28/20 at 6:22 pm
Posted on 12/28/20 at 6:24 pm to TigerDentist
quote:
Also 2018 included a decent amount of JUCO and grad transfers who only have a year or two of eligibility.
Yeah, the JUCOs and grad transfers are another huge part of our roster hole.
OTOH, I would say those guys worked out pretty well.
It’s an interesting strategy that we seem to be testing again for the 2021 class - getting those experienced guys can obviously be very effective at filling needs short term. See: Damien Lewis, Jabril Cox, and Cole Tracy. Those guys can also win you a championship. See: Joe Burrow. But that strategy does put you in a position of constantly having to rebuild, which could lead to more boom/bust types of seasons.
Posted on 12/28/20 at 6:27 pm to lostinbr
Playing so many Soph & true Freshmen in 2020, means the 2021 / 2022 classes are can’t miss to not have a huge drop off again. And CEO has to get every draft eligible player, who isn’t 1st round material to come back.
Posted on 12/28/20 at 6:33 pm to lostinbr
If data is an art form then this poster just painted his masterpiece!
Posted on 12/28/20 at 6:36 pm to lostinbr
Great work man! Uber impressive
Page 1 of 4
Popular
Back to top
Follow TigerDroppings for LSU Football News